|
Post by Blackdawg on Jan 23, 2019 1:00:20 GMT -6
Anyone here tried the sparkos Lab pro opamp? On paper looks crazy good for clean mastering stuff or super detailed for a mic preamp
|
|
|
Post by matt@IAA on Jan 23, 2019 1:44:29 GMT -6
Interesting reading. Lots of fancy words. The maximum gain is not relevant unless you’re doing DC gain for some reason. The number that matters is the max gain through the audio band, which is about 75 or 80 dB. But that being said what audio application needs a single op amp to put out 80 dB? Even a high gain mic pre usually won’t ask for more than 40 from the op amp.
Edited to say - and this makes sense - he suggests that open loop gain versus demand (gain margin I suppose) is directly proportional to THD. I have never seen this written anywhere in those terms, but I suppose it may be true.
The slew rates are also interesting - the positive and negative slew rates are different (20 up 12 down). Not sure how much this matters.
Buy one and try it out!
Reading through the white paper has me cracking up. The guy that wrote it is hilarious.
|
|
ericn
Temp
Balance Engineer
Posts: 15,012
Member is Online
|
Post by ericn on Jan 23, 2019 7:48:24 GMT -6
Guys like this seam to come and go in the Audiophile world on an almost yearly basis. Do they sound better ? I don’t know, the complaint I always hear is most of these guys are one opamp design fits all philosophy and from the few guys I know in that world that know what they are doing always seam to say like so many things it depends on the application.
|
|
|
Post by Blackdawg on Jan 23, 2019 8:56:11 GMT -6
Well idk if I buy it. But I might grab a couple just for fun. The 2520 footprint one that is. Be interesting to compare to app992 op amps perhaps and some je990s
I also like the 2520 to DIP8 adapter board. Pretty cool.
|
|
|
Post by Blackdawg on Jan 23, 2019 11:50:05 GMT -6
|
|
|
Post by matt@IAA on Jan 23, 2019 15:06:51 GMT -6
I never considered a lot of the things that the SparkoS whitepaper mentions when comparing monolithic op amps to discrete designs. One thing I took for granted was the compensation scheme. None of the discrete designs I have seen use single pole compensation. Just for funsies, it is interesting to have a look at some simulation results for open loop gain. All of these graphs are done in LTSpice based on my understanding of the schematics. Neumann OA10 - with compensation components for unity gain stability API 2520 QE AM10 M2:
|
|
|
Post by Blackdawg on Jan 23, 2019 15:31:36 GMT -6
I like where this is going..
but I need more info on what Im looking at here. What line means what exactly?
|
|
|
Post by matt@IAA on Jan 23, 2019 15:50:34 GMT -6
Open loop gain is when the op amp has no feedback applied. When you close the loop (connect out to -IN) you get closed loop gain. The resistors and such that you put around it set the closed loop gain.
So this is the open loop gain curve for these op amps. They are frequency limited by parasitic capacitance in the transistors themselves - they actually have a "speed". So eventually the op amp has less and less gain at higher and higher frequencies.
The gain bandwidth product (GBW) is gain * frequency. For single pole op amps like on most IC chips since the rolloff is constant, you can take any point on the curve and multiple the gain by the frequency to know how much gain you have at that point. These designs have a nonlinear rolloff, so the GBW is not constant. Basically though where the gain line hits 0 is the maximum frequency the op amp can give any gain to - after that, it'll actually attenuate. This is the bandwith.
The dotted line is phase. Op amps have compensation for stability inside, and this plus the internal parasitic capacitance cause a phase shift through the op amp. When the op amp hits zero gain, you had better have a phase angle of greater than -180 degrees, because you take the op amp feedback and add it into the -IN terminal, which is another 180 phase change. In other words, if you're south of -180, that feedback just became feedforward and you now have an oscillator.
The three lines I drew are the DC gain, the audio band max gain (if you asked the op amp for more gain than that, you'd start to roll off the highs), and the phase margin at 0 gain (the distance between the phase angle and -180).
Phase margin also is important because less than 60 degrees phase margin will cause some ringing, oscillation, overshoot, whatever on a square wave impulse. It can also cause over-amplification of highs (that's really what a square wave response is showing you - infinite bandwidth response). Depending on how much gain you ask of the op amp and how you use it, that may be bad. Capacitive loads and the source resistance seen by your input terminals can also chew up phase margin.
The Sparkos guy says that the THD an op amp produces is less the more gain margin it has (i.e., if you're asking it for 20dB and it can make 80dB at that frequency, you'd have 60 dB margin). I'd never compared the THD and gain margin, but holds true. So then, look at the gain at 20 kHz, and maybe you have a reasonable analog for THD (higher gain = less THD).
|
|
|
Post by Blackdawg on Jan 23, 2019 16:12:42 GMT -6
Great stuff man!
So going off that, the API2520 is the "best" opamp. Excluding other factors of course.
|
|
|
Post by matt@IAA on Jan 23, 2019 16:24:53 GMT -6
The M2 has higher gain at 20k and a teensy bit more phase margin.
|
|
|
Post by matt@IAA on Jan 23, 2019 16:48:47 GMT -6
Other things to consider - class A output capability. The 2520 has around 16 mA, I think. The M2 has around 4. The Neumann has 2.4. The AM10 has around 3.5. This determines when the op amp transitions from class A to class AB mode, meaning you start to get more and more odd order harmonics. Symmetry (AB, push pull) suppress even harmonics, so you get more odd harmonics, assymmetry (class A) has higher magnitude even harmonics.
Even harmonics sound warm, thick, fat. Odd sound bright or detailed. Both can have too much. Some people prefer one or another.
So even here it hard to say "best". For starters, some people really dig the Neumann, but you can push it into class AB very easily. And even that doesn't tell the whole story...the FFT tells the tale. When I fiddle around with these on a test bench, the 2520 winds up with a lot of odd harmonic content. Probably why it has the reputation for being "bright" and "detailed". So it comes down to minor (like less than .001% THD) differences in total distortion, and different distributions of the THD harmonic by harmonic.
But you can also somewhat control this by the load you present the op amp. If you go into a 10k stage, you'll ask very little current, so it's no big deal to stay in class A. But if you hit something with a 600 ohm input, suddenly you might be sucking down some current! Especially if you're going through a 1:2 output transformer...
|
|
|
Post by Omicron9 on Jan 24, 2019 10:32:07 GMT -6
Anyone here tried the sparkos Lab pro opamp? On paper looks crazy good for clean mastering stuff or super detailed for a mic preamp Is this the one under discussion: Pro Discrete Op Amp - SparkoS?
|
|
|
Post by Blackdawg on Jan 24, 2019 10:38:21 GMT -6
Anyone here tried the sparkos Lab pro opamp? On paper looks crazy good for clean mastering stuff or super detailed for a mic preamp Is this the one under discussion: Pro Discrete Op Amp - SparkoS? Yes. Well and the other smaller ones they make for the DIP8 ones. But mostly the pro one.
|
|
|
Post by Blackdawg on Jan 25, 2019 11:09:11 GMT -6
Well i ordered two of these to add into my opamp shoot out. Should be cool to hear.
|
|
|
Post by EmRR on Jan 25, 2019 13:10:15 GMT -6
Interesting reading. Lots of fancy words. The maximum gain is not relevant unless you’re doing DC gain for some reason. The number that matters is the max gain through the audio band, which is about 75 or 80 dB. But that being said what audio application needs a single op amp to put out 80 dB? Even a high gain mic pre usually won’t ask for more than 40 from the op amp. Edited to say - and this makes sense - he suggests that open loop gain versus demand (gain margin I suppose) is directly proportional to THD. I have never seen this written anywhere in those terms, but I suppose it may be true. Yeah it's important for people to realize that most audio products don't push op amps past 40dB. Sure, the more feedback you have, the lower THD is. Feedback changes the harmonic profile, creating more and more higher order harmonics, so it has to be shoved down 40+dB to overcome secondary by-products. That's kind of the point of op amp design theory, and open loop gain tells you how much feedback gain squashing you can do. Conversely, the earliest op amps were mainly for DC amplification purposes. I've got a stock of one type that requires typical mic preamp feedback amounts to still be flat to 10kHz, and there's not much past that.
|
|
|
Post by hugostiglitz on Jan 26, 2019 8:38:47 GMT -6
Off-Topic:
I found the Sparkos as well as the Sonic Imagery Labs op-amps quite expensive. Is there anything DIY which offer the same form factor of discrete op-amps in a drip8 format?
|
|
|
Post by Blackdawg on Jan 26, 2019 12:44:44 GMT -6
Off-Topic: I found the Sparkos as well as the Sonic Imagery Labs op-amps quite expensive. Is there anything DIY which offer the same form factor of discrete op-amps in a drip8 format? I've been talking with Pier who makes the APP992. Which is a diy 2520 opamp. But he also makes it as a DIP8 single and dual. I didn't get a price though yet. But would expect it to be 20-30 bucks each. Which I'm a bit agast about as getting a batch with the smds on them can't cost much. Still. Those are the only discreet DIP8 opamos I've found so far. Not that I'm looking super hard. I think I'll end up with the app992DIP8s to try in my sontec clone build.
|
|
|
Post by matt@IAA on Jan 26, 2019 13:17:32 GMT -6
Off-Topic: I found the Sparkos as well as the Sonic Imagery Labs op-amps quite expensive. Is there anything DIY which offer the same form factor of discrete op-amps in a drip8 format? I've been talking with Pier who makes the APP992. Which is a diy 2520 opamp. But he also makes it as a DIP8 single and dual. I didn't get a price though yet. But would expect it to be 20-30 bucks each. Which I'm a bit agast about as getting a batch with the smds on them can't cost much. Still. Those are the only discreet DIP8 opamos I've found so far. Not that I'm looking super hard. I think I'll end up with the app992DIP8s to try in my sontec clone build. Wish the APP992 had a data sheet. It’d be interesting to see some more technical info. I can re-footprint any op amp you want into a DIP8 footprint. PCBs that size are cheap (like $15 for 90). As long as there’s real estate for the larger pcb to “hover”. Then you could solder em up yourself. Or make a 2520 to DIp8 adapter, also very simple.
|
|
|
Post by Blackdawg on Jan 26, 2019 13:21:13 GMT -6
I've been talking with Pier who makes the APP992. Which is a diy 2520 opamp. But he also makes it as a DIP8 single and dual. I didn't get a price though yet. But would expect it to be 20-30 bucks each. Which I'm a bit agast about as getting a batch with the smds on them can't cost much. Still. Those are the only discreet DIP8 opamos I've found so far. Not that I'm looking super hard. I think I'll end up with the app992DIP8s to try in my sontec clone build. Wish the APP992 had a data sheet. It’d be interesting to see some more technical info. I can re-footprint any op amp you want into a DIP8 footprint. PCBs that size are cheap (like $15 for 90). As long as there’s real estate for the larger pcb to “hover”. Then you could solder em up yourself. Or make a 2520 to DIp8 adapter, also very simple. Sparkos makes a nice 2520 to dip adapter. Problem with that is it's only applicable to single opamps not dual. I'm excited to get the app992. It has a super fast slew rate. Likely end up in my sontec build but be fun to demo with all the others I have. So far I think I have 10 now.
|
|
|
Post by matt@IAA on Jan 26, 2019 13:33:24 GMT -6
Yeah, there’s not many dual op amps in the 2520 footprint so an adapter wouldn’t do you much.
In my understanding there’s no benefit to super duper ultra high slew rates. To take full advantage of a 75 V/uS slew rate you’d have to be pushing 150V p-p! A rule of thumb is that slew rate needs to be minimum .5 V/us per volt of peak output to prevent audible distortion. There’s really no benefit to anything much higher than 8 or 10 V/uS. But, I could be wrong.
|
|
|
Post by Blackdawg on Jan 26, 2019 13:36:05 GMT -6
Yeah, there’s not many dual op amps in the 2520 footprint so an adapter wouldn’t do you much. In my understanding there’s no benefit to super duper ultra high slew rates. To take full advantage of a 75 V/uS slew rate you’d have to be pushing 150V p-p! A rule of thumb is that slew rate needs to be minimum .5 V/us per volt of peak output to prevent audible distortion. There’s really no benefit to anything much higher than 8 or 10 V/uS. But, I could be wrong. I have no idea either. I do know he says it has a 75v/uS slew rate though. It's supposed to be a very detailed opamps. Similar to a 990 but tweaked
|
|
|
Post by matt@IAA on Jan 26, 2019 13:38:56 GMT -6
Yeah, there’s not many dual op amps in the 2520 footprint so an adapter wouldn’t do you much. In my understanding there’s no benefit to super duper ultra high slew rates. To take full advantage of a 75 V/uS slew rate you’d have to be pushing 150V p-p! A rule of thumb is that slew rate needs to be minimum .5 V/us per volt of peak output to prevent audible distortion. There’s really no benefit to anything much higher than 8 or 10 V/uS. But, I could be wrong. I have no idea either. I do know he says it has a 75v/uS slew rate though. It's supposed to be a very detailed opamps. Similar to a 990 but tweaked A 990 only has a slew rate of 18v/us. Also, I think the APP is using jfet inputs.
|
|
|
Post by Blackdawg on Jan 26, 2019 13:46:34 GMT -6
I have no idea either. I do know he says it has a 75v/uS slew rate though. It's supposed to be a very detailed opamps. Similar to a 990 but tweaked A 990 only has a slew rate of 18v/us. Also, I think the APP is using jfet inputs. Yeah it is jfet. That's just what I've read. Never seen a full spec sheet.
|
|